Novel MSCNet Yields 81.24% Dice Score for Coronary Artery Segmentation on CTA

MSCNet, a deep learning network for coronary artery segmentation on CTA, achieved a Dice score of 81.24%—outperforming prior models by 3.5–3.9 percentage points—in a retrospective evaluation on the ImageCAS dataset.

• Retrospective development study using the ImageCAS dataset; sample size and patient demographics not reported.
• No external or prospective validation performed; generalizability unknown.
• Model incorporates multi-scale cascade encoding and dynamic spatial context enhancement to address segmentation breaks and false positives.

RadPigeon summaries are original and for information only. They are not clinical advice.